Merge branch 'dev'v6.66

22 files changed, 143 insertions, 49 deletions

diff --git a/CMakeLists.txt b/CMakeLists.txt
index d786752..1d91b0e 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 2.6...3.0.2)
 
-project(xbyak LANGUAGES CXX VERSION 6.65)
+project(xbyak LANGUAGES CXX VERSION 6.66)
 
 file(GLOB headers xbyak/*.h)
 
diff --git a/doc/changelog.md b/doc/changelog.md
index b976c90..5e6fa9a 100644
--- a/doc/changelog.md
+++ b/doc/changelog.md
@@ -1,5 +1,6 @@
 # History
 
+* 2022/Nov/25 ver 6.66 support RAO-INT
 * 2022/Nov/22 ver 6.65 consider x32
 * 2022/Nov/04 ver 6.64 some vmov* support addressing with mask
 * 2022/Oct/06 ver 6.63 vpmadd52{h,l}uq support AVX-IFMA
diff --git a/gen/Makefile b/gen/Makefile
index 97a6846..f254d71 100644
--- a/gen/Makefile
+++ b/gen/Makefile
@@ -1,6 +1,6 @@
 TARGET=../xbyak/xbyak_mnemonic.h
 BIN=sortline gen_code gen_avx512
-CFLAGS=-I../ -O2 -DXBYAK_NO_OP_NAMES -Wall -Wextra -Wno-missing-field-initializers $(CXXFLAGS) $(CPPFLAGS) $(LDFLAGS)
+CFLAGS=-I../ -O2 -Wall -Wextra -Wno-missing-field-initializers $(CXXFLAGS) $(CPPFLAGS) $(LDFLAGS)
 all: $(TARGET) ../CMakeLists.txt ../meson.build ../readme.md ../readme.txt
 sortline: sortline.cpp
 	$(CXX) $(CFLAGS) $< -o $@
diff --git a/gen/gen_code.cpp b/gen/gen_code.cpp
index b89009b..7a5c575 100644
--- a/gen/gen_code.cpp
+++ b/gen/gen_code.cpp
@@ -807,6 +807,23 @@ void put()
 			printf("void %s(const Operand& op, uint32_t imm) { opRM_I(op, imm, 0x%02X, %d); }\n", p->name, p->code, p->ext);
 		}
 	}
+	{
+		const struct Tbl {
+			const char *name;
+			uint8_t prefix;
+		} tbl[] = {
+			{ "aadd", 0 },
+			{ "aand", 0x66 },
+			{ "aor", 0xF2 },
+			{ "axor", 0xF3 },
+		};
+		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+			const Tbl *p = &tbl[i];
+			printf("void %s(const Address& addr, const Reg32e &reg) { ", p->name);
+			if (p->prefix) printf("db(0x%02X); ", p->prefix);
+			printf("opModM(addr, reg, 0x0F, 0x38, 0x0FC); }\n");
+		}
+	}
 
 	{
 		const struct Tbl {
diff --git a/meson.build b/meson.build
index 971126b..f296532 100644
--- a/meson.build
+++ b/meson.build
@@ -5,7 +5,7 @@
 project(
 	'xbyak',
 	'cpp',
-	version: '6.65',
+	version: '6.66',
 	license: 'BSD-3-Clause',
 	default_options: 'b_ndebug=if-release'
 )
diff --git a/readme.md b/readme.md
index 5052589..0fa16fa 100644
--- a/readme.md
+++ b/readme.md
@@ -1,5 +1,5 @@
 
-# Xbyak 6.65 [![Badge Build]][Build Status]
+# Xbyak 6.66 [![Badge Build]][Build Status]
 
 *A C++ JIT assembler for x86 (IA32), x64 (AMD64, x86-64)*
 
diff --git a/readme.txt b/readme.txt
index f2eedb6..ed1f2dd 100644
--- a/readme.txt
+++ b/readme.txt
@@ -1,5 +1,5 @@
 

-    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 6.65

+    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 6.66

 

 -----------------------------------------------------------------------------

 ◎概要

@@ -402,6 +402,7 @@ sample/{echo,hello}.bfは http://www.kmonos.net/alang/etc/brainfuck.php から
 -----------------------------------------------------------------------------

 ◎履歴

 

+2022/11/25 ver 6.66 RAO-INTサポート

 2022/11/22 ver 6.65 x32動作確認

 2022/11/04 ver 6.64 vmov*命令をmaskつきアドレッシング対応修正

 2022/10/06 ver 6.63 AVX-IFMA用のvpmadd52{h,l}uq対応

diff --git a/sample/Makefile b/sample/Makefile
index 7c910bb..7e8ab7b 100644
--- a/sample/Makefile
+++ b/sample/Makefile
@@ -1,6 +1,9 @@
 XBYAK_INC=../xbyak/xbyak.h
+CXX?=g++
 
-BOOST_EXIST=$(shell echo "\#include <boost/spirit/core.hpp>" | (gcc -E - 2>/dev/null) | grep "boost/spirit/core.hpp" >/dev/null && echo "1")
+#BOOST_EXIST=$(shell echo "#include <boost/spirit/core.hpp>" | $CXX -x c++ -c - 2>/dev/null && echo 1)
+# I don't know why the above code causes an error on GitHub action.
+BOOST_EXIST?=0
 UNAME_M=$(shell uname -m)
 
 ONLY_64BIT=0
@@ -104,7 +107,7 @@ profiler-vtune: profiler.cpp ../xbyak/xbyak_util.h
 	$(CXX) $(CFLAGS) profiler.cpp -o $@ -DXBYAK_USE_VTUNE -I /opt/intel/vtune_amplifier/include/ -L /opt/intel/vtune_amplifier/lib64 -ljitprofiling -ldl
 
 clean:
-	rm -rf *.o $(TARGET) *.exe profiler profiler-vtune
+	rm -rf $(TARGET) profiler profiler-vtune
 
 test : test0.cpp $(XBYAK_INC)
 test64: test0.cpp $(XBYAK_INC)
diff --git a/sample/quantize.cpp b/sample/quantize.cpp
index 6bdf0d0..ba0fd22 100644
--- a/sample/quantize.cpp
+++ b/sample/quantize.cpp
@@ -199,7 +199,7 @@ int main(int argc, char *argv[])
 		quantize2(dest2, src, qTbl);
 		for (int i = 0; i < N; i++) {
 			if (dest[i] != dest2[i]) {
-				printf("err[%d] %d %d\n", i, dest[i], dest2[i]);
+				printf("err[%d] %u %u\n", i, dest[i], dest2[i]);
 			}
 		}
 
diff --git a/sample/test_util.cpp b/sample/test_util.cpp
index a20d2df..ef6e3fa 100644
--- a/sample/test_util.cpp
+++ b/sample/test_util.cpp
@@ -93,6 +93,7 @@ void putCPUinfo(bool onlyCpuidFeature)
 		{ Cpu::tAVX_VNNI_INT8, "avx_vnni_int8" },
 		{ Cpu::tAVX_NE_CONVERT, "avx_ne_convert" },
 		{ Cpu::tAVX_IFMA, "avx_ifma" },
+		{ Cpu::tRAO_INT, "rao-int" },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		if (cpu.has(tbl[i].type)) printf(" %s", tbl[i].str);
diff --git a/sample/toyvm.cpp b/sample/toyvm.cpp
index 1e558ff..dff0cb7 100644
--- a/sample/toyvm.cpp
+++ b/sample/toyvm.cpp
@@ -5,8 +5,8 @@
 
 	mem_  4byte x 65536
 
-	���٤Ƥ�̿���4byte����
-	¨�ͤ�����16bit
+	all instructions are fixed at 4 bytes.
+	all immediate values are 16-bit.
 
 	R = A or B
 	vldiR, imm  ; R = imm
@@ -109,7 +109,7 @@ public:
 				reg[r] -= imm;
 				break;
 			case PUT:
-				printf("%c %8d(0x%08x)\n", 'A' + r, reg[r], reg[r]);
+				printf("%c %8u(0x%08x)\n", 'A' + r, reg[r], reg[r]);
 				break;
 			case JNZ:
 				if (reg[r] != 0) pc += static_cast<signed short>(imm);
@@ -294,7 +294,7 @@ lp:
 	p = t;
 	n--;
 	if (n != 0) goto lp;
-	printf("c=%d(0x%08x)\n", c, c);
+	printf("c=%u(0x%08x)\n", c, c);
 }
 
 int main()
diff --git a/test/Makefile b/test/Makefile
index 2b9bd1a..feef445 100644
--- a/test/Makefile
+++ b/test/Makefile
@@ -23,9 +23,9 @@ endif
 
 all: $(TARGET)
 
-CFLAGS_WARN=-Wall -Wextra -Wformat=2 -Wcast-qual -Wcast-align -Wwrite-strings -Wfloat-equal -Wpointer-arith
+CFLAGS_WARN=-Wall -Wextra -Wformat=2 -Wcast-qual -Wwrite-strings -Wfloat-equal -Wpointer-arith
 
-CFLAGS=-O2 -fomit-frame-pointer -Wall -fno-operator-names -I../ -I./ $(CFLAGS_WARN) $(CXXFLAGS) $(CPPFLAGS) $(LDFLAGS) #-std=c++0x
+CFLAGS=-O2 -Wall -I../ -I./ $(CFLAGS_WARN) $(CXXFLAGS) $(CPPFLAGS) $(LDFLAGS) #-std=c++0x
 make_nm:
 	$(CXX) $(CFLAGS) make_nm.cpp -o $@
 normalize_prefix: normalize_prefix.cpp ../xbyak/xbyak.h
@@ -56,12 +56,11 @@ noexception: noexception.cpp ../xbyak/xbyak.h
 test_nm: normalize_prefix $(TARGET)
 	$(MAKE) -C ../gen
 ifneq ($(ONLY_64BIT),1)
-	./test_nm.sh
-	./test_nm.sh noexcept
-	./noexception
-	./test_nm.sh Y
-	./test_nm.sh avx512
-	./test_address.sh
+	env CXX=$(CXX) sh -e ./test_nm.sh
+	env CXX=$(CXX) sh -e ./test_nm.sh noexcept
+	env CXX=$(CXX) sh -e ./test_nm.sh Y
+	env CXX=$(CXX) sh -e ./test_nm.sh avx512
+	env CXX=$(CXX) sh -e ./test_address.sh
 	./jmp
 	./cvt_test32
 endif
@@ -70,32 +69,32 @@ endif
 	./misc32
 	./cvt_test
 ifeq ($(BIT),64)
-	./test_address.sh 64
+	env CXX=$(CXX) sh -e ./test_address.sh 64
 ifneq ($(X32),1)
-	./test_nm.sh 64
-	./test_nm.sh Y64
+	env CXX=$(CXX) sh -e ./test_nm.sh 64
+	env CXX=$(CXX) sh -e ./test_nm.sh Y64
 endif
 	./jmp64
 endif
 
 test_avx: normalize_prefix
 ifneq ($(ONLY_64BIT),0)
-	./test_avx.sh
-	./test_avx.sh Y
+	env CXX=$(CXX) sh -e ./test_avx.sh
+	env CXX=$(CXX) sh -e ./test_avx.sh Y
 endif
 ifeq ($(BIT),64)
-	./test_avx.sh 64
+	env CXX=$(CXX) sh -e ./test_avx.sh 64
 ifneq ($(X32),1)
-	./test_avx.sh Y64
+	env CXX=$(CXX) sh -e ./test_avx.sh Y64
 endif
 endif
 
 test_avx512: normalize_prefix
 ifneq ($(ONLY_64BIT),0)
-	./test_avx512.sh
+	env CXX=$(CXX) sh -e ./test_avx512.sh
 endif
 ifeq ($(BIT),64)
-	./test_avx512.sh 64
+	env CXX=$(CXX) sh -e ./test_avx512.sh 64
 endif
 
 detect_x32: detect_x32.c
diff --git a/test/make_nm.cpp b/test/make_nm.cpp
index 5106bf9..e5939eb 100644
--- a/test/make_nm.cpp
+++ b/test/make_nm.cpp
@@ -1018,9 +1018,7 @@ class Test {
 	}
 	void putCmov() const
 	{
-		const struct {
-			const char *s;
-		} tbl[] = {
+		const char tbl[][4] = {
 			"o",
 			"no",
 			"b",
@@ -1054,11 +1052,11 @@ class Test {
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			char buf[32];
-			snprintf(buf, sizeof(buf), "cmov%s", tbl[i].s);
+			snprintf(buf, sizeof(buf), "cmov%s", tbl[i]);
 			put(buf, REG16, REG16|MEM);
 			put(buf, REG32, REG32|MEM);
 			put(buf, REG64, REG64|MEM);
-			snprintf(buf, sizeof(buf), "set%s", tbl[i].s);
+			snprintf(buf, sizeof(buf), "set%s", tbl[i]);
 			put(buf, REG8|REG8_3|MEM);
 		}
 	}
diff --git a/test/misc.cpp b/test/misc.cpp
index 6225690..7653673 100644
--- a/test/misc.cpp
+++ b/test/misc.cpp
@@ -1993,3 +1993,66 @@ CYBOZU_TEST_AUTO(minmax)
 	CYBOZU_TEST_EQUAL((std::min)(3, 4), local::min_(3, 4));
 	CYBOZU_TEST_EQUAL((std::max)(3, 4), local::max_(3, 4));
 }
+
+CYBOZU_TEST_AUTO(rao_int)
+{
+	struct Code : Xbyak::CodeGenerator {
+		Code()
+		{
+#ifdef XBYAK64
+			aadd(ptr[rax], ecx);
+			aadd(ptr[eax], ecx);
+			aadd(ptr[rax], r10);
+			aand(ptr[rax], ecx);
+			aand(ptr[eax], ecx);
+			aand(ptr[rax], r10);
+			aor(ptr[rax], ecx);
+			aor(ptr[eax], ecx);
+			aor(ptr[rax], r10);
+			axor(ptr[rax], ecx);
+			axor(ptr[eax], ecx);
+			axor(ptr[rax], r10);
+#else
+			aadd(ptr[eax], ecx);
+			aand(ptr[eax], ecx);
+			aor(ptr[eax], ecx);
+			axor(ptr[eax], ecx);
+#endif
+		}
+	} c;
+	const uint8_t tbl[] = {
+#ifdef XBYAK64
+		// aadd
+		0x0f, 0x38, 0xfc, 0x08,
+		0x67, 0x0f, 0x38, 0xfc, 0x08,
+		0x4c, 0x0f, 0x38, 0xfc, 0x10,
+
+		// aand
+		0x66, 0x0f, 0x38, 0xfc, 0x08,
+		0x66, 0x67, 0x0f, 0x38, 0xfc, 0x08,
+		0x66, 0x4c, 0x0f, 0x38, 0xfc, 0x10,
+
+		// aor
+		0xf2, 0x0f, 0x38, 0xfc, 0x08,
+		0xf2, 0x67, 0x0f, 0x38, 0xfc, 0x08,
+		0xf2, 0x4c, 0x0f, 0x38, 0xfc, 0x10,
+
+		// axor
+		0xf3, 0x0f, 0x38, 0xfc, 0x08,
+		0xf3, 0x67, 0x0f, 0x38, 0xfc, 0x08,
+		0xf3, 0x4c, 0x0f, 0x38, 0xfc, 0x10,
+#else
+		// aadd
+		0x0f, 0x38, 0xfc, 0x08,
+		// aand
+		0x66, 0x0f, 0x38, 0xfc, 0x08,
+		// aor
+		0xf2, 0x0f, 0x38, 0xfc, 0x08,
+		// axor
+		0xf3, 0x0f, 0x38, 0xfc, 0x08,
+#endif
+	};
+	const size_t n = sizeof(tbl) / sizeof(tbl[0]);
+	CYBOZU_TEST_EQUAL(c.getSize(), n);
+	CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
+}
diff --git a/test/noexception.cpp b/test/noexception.cpp
index 04a6dbc..9ef0ee8 100644
--- a/test/noexception.cpp
+++ b/test/noexception.cpp
@@ -56,7 +56,7 @@ void test2()
 void test3()
 {
 	static struct EmptyAllocator : Xbyak::Allocator {
-		uint8_t *alloc() { return 0; }
+		uint8_t *alloc(size_t) { return 0; }
 	} emptyAllocator;
 	struct Code : CodeGenerator {
 		Code() : CodeGenerator(8, 0, &emptyAllocator)
diff --git a/test/test_address.sh b/test/test_address.sh
index d283a5f..7960700 100755
--- a/test/test_address.sh
+++ b/test/test_address.sh
@@ -5,9 +5,11 @@ FILTER="grep -v warning"
 sub()
 {
 
-CFLAGS="-Wall -fno-operator-names -I../ $OPT2"
+CFLAGS="-Wall -I../ $OPT2"
+CXX=${CXX:=g++}
+
 echo "compile address.cpp"
-g++ $CFLAGS address.cpp -o address
+$CXX $CFLAGS address.cpp -o address
 
 ./address $1 > a.asm
 echo "asm"
@@ -17,7 +19,7 @@ awk '{printf "%s", sub(/-$/, "", $3) ? $3 : $3 ORS}' a.lst | $FILTER > ok.lst
 echo "xbyak"
 ./address $1 jit > nm.cpp
 echo "compile nm_frame.cpp"
-g++ $CFLAGS -DXBYAK_TEST nm_frame.cpp -o nm_frame
+$CXX $CFLAGS -DXBYAK_TEST nm_frame.cpp -o nm_frame
 ./nm_frame > x.lst
 diff ok.lst x.lst && echo "ok"
 
diff --git a/test/test_avx.sh b/test/test_avx.sh
index 34dc1e5..35087cd 100755
--- a/test/test_avx.sh
+++ b/test/test_avx.sh
@@ -1,6 +1,7 @@
 #!/bin/sh
 
 FILTER="grep -v warning"
+CXX=${CXX:=g++}
 
 case $1 in
 Y)
@@ -31,9 +32,9 @@ Y64)
 	;;
 esac
 
-CFLAGS="-Wall -fno-operator-names -I../ $OPT2 -DUSE_AVX"
+CFLAGS="-Wall -I../ $OPT2 -DUSE_AVX"
 echo "compile make_nm.cpp"
-g++ $CFLAGS make_nm.cpp -o make_nm
+$CXX $CFLAGS make_nm.cpp -o make_nm
 
 ./make_nm > a.asm
 echo "asm"
@@ -43,6 +44,6 @@ awk '$3 != "1+1" {printf "%s", sub(/-$/, "", $3) ? $3 : $3 ORS}' a.lst | $FILTER
 echo "xbyak"
 ./make_nm jit > nm.cpp
 echo "compile nm_frame.cpp"
-g++ $CFLAGS -DXBYAK_TEST nm_frame.cpp -o nm_frame
+$CXX $CFLAGS -DXBYAK_TEST nm_frame.cpp -o nm_frame
 ./nm_frame | $FILTER > x.lst
 diff -B ok.lst x.lst && echo "ok"
diff --git a/test/test_avx512.sh b/test/test_avx512.sh
index 17edfee..90d14df 100755
--- a/test/test_avx512.sh
+++ b/test/test_avx512.sh
@@ -1,6 +1,7 @@
 #!/bin/sh
 
 FILTER="grep -v warning"
+CXX=${CXX:=g++}
 
 case $1 in
 64)
@@ -18,9 +19,9 @@ case $1 in
 	;;
 esac
 
-CFLAGS="-Wall -fno-operator-names -I../ $OPT2 -DUSE_AVX512"
+CFLAGS="-Wall -I../ $OPT2 -DUSE_AVX512"
 echo "compile make_512.cpp"
-g++ $CFLAGS make_512.cpp -o make_512
+$CXX $CFLAGS make_512.cpp -o make_512
 
 ./make_512 > a.asm
 echo "asm"
@@ -30,6 +31,6 @@ awk '{printf "%s", sub(/-$/, "", $3) ? $3 : $3 ORS}' a.lst | $FILTER > ok.lst
 echo "xbyak"
 ./make_512 jit > nm.cpp
 echo "compile nm_frame.cpp"
-g++ $CFLAGS -DXBYAK_TEST nm_frame.cpp -o nm_frame -DXBYAK_AVX512
+$CXX $CFLAGS -DXBYAK_TEST nm_frame.cpp -o nm_frame -DXBYAK_AVX512
 ./nm_frame | $FILTER > x.lst
 diff -B ok.lst x.lst && echo "ok"
diff --git a/test/test_nm.sh b/test/test_nm.sh
index afa2b1e..019f278 100755
--- a/test/test_nm.sh
+++ b/test/test_nm.sh
@@ -1,6 +1,7 @@
 #!/bin/sh
 
 FILTER=cat
+CXX=${CXX:=g++}
 
 case $1 in
 Y)
@@ -44,9 +45,9 @@ noexcept)
 	;;
 esac
 
-CFLAGS="-Wall -fno-operator-names -I../ $OPT2"
+CFLAGS="-Wall -I../ $OPT2"
 echo "compile make_nm.cpp with $CFLAGS"
-g++ $CFLAGS make_nm.cpp -o make_nm
+$CXX $CFLAGS make_nm.cpp -o make_nm
 
 ./make_nm > a.asm
 echo "asm"
@@ -56,6 +57,6 @@ awk '$3 != "1+1" {printf "%s", sub(/-$/, "", $3) ? $3 : $3 ORS}' a.lst | $FILTER
 echo "xbyak"
 ./make_nm jit > nm.cpp
 echo "compile nm_frame.cpp"
-g++ $CFLAGS -DXBYAK_TEST nm_frame.cpp -o nm_frame
+$CXX $CFLAGS -DXBYAK_TEST nm_frame.cpp -o nm_frame
 ./nm_frame | $FILTER > x.lst
 diff -B ok.lst x.lst && echo "ok"
diff --git a/xbyak/xbyak.h b/xbyak/xbyak.h
index 2a9c031..86372a0 100644
--- a/xbyak/xbyak.h
+++ b/xbyak/xbyak.h
@@ -155,7 +155,7 @@ namespace Xbyak {
 
 enum {
 	DEFAULT_MAX_CODE_SIZE = 4096,
-	VERSION = 0x6650 /* 0xABCD = A.BC(.D) */
+	VERSION = 0x6660 /* 0xABCD = A.BC(.D) */
 };
 
 #ifndef MIE_INTEGER_TYPE_DEFINED
diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h
index 33bd5c0..13e52e1 100644
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@@ -1,4 +1,6 @@
-const char *getVersionString() const { return "6.65"; }
+const char *getVersionString() const { return "6.66"; }
+void aadd(const Address& addr, const Reg32e &reg) { opModM(addr, reg, 0x0F, 0x38, 0x0FC); }
+void aand(const Address& addr, const Reg32e &reg) { db(0x66); opModM(addr, reg, 0x0F, 0x38, 0x0FC); }
 void adc(const Operand& op, uint32_t imm) { opRM_I(op, imm, 0x10, 2); }
 void adc(const Operand& op1, const Operand& op2) { opRM_RM(op1, op2, 0x10); }
 void adcx(const Reg32e& reg, const Operand& op) { opGen(reg, op, 0xF6, 0x66, isREG32_REG32orMEM, NONE, 0x38); }
@@ -24,6 +26,8 @@ void andnpd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x55, 0x66, isXM
 void andnps(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x55, 0x100, isXMM_XMMorMEM); }
 void andpd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x54, 0x66, isXMM_XMMorMEM); }
 void andps(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x54, 0x100, isXMM_XMMorMEM); }
+void aor(const Address& addr, const Reg32e &reg) { db(0xF2); opModM(addr, reg, 0x0F, 0x38, 0x0FC); }
+void axor(const Address& addr, const Reg32e &reg) { db(0xF3); opModM(addr, reg, 0x0F, 0x38, 0x0FC); }
 void bextr(const Reg32e& r1, const Operand& op, const Reg32e& r2) { opGpr(r1, op, r2, T_0F38, 0xf7, false); }
 void blendpd(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0x0D, 0x66, isXMM_XMMorMEM, static_cast<uint8_t>(imm), 0x3A); }
 void blendps(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0x0C, 0x66, isXMM_XMMorMEM, static_cast<uint8_t>(imm), 0x3A); }
diff --git a/xbyak/xbyak_util.h b/xbyak/xbyak_util.h
index 8431925..e1e4476 100644
--- a/xbyak/xbyak_util.h
+++ b/xbyak/xbyak_util.h
@@ -414,6 +414,7 @@ public:
 	XBYAK_DEFINE_TYPE(69, tAVX_VNNI_INT8);
 	XBYAK_DEFINE_TYPE(70, tAVX_NE_CONVERT);
 	XBYAK_DEFINE_TYPE(71, tAVX_IFMA);
+	XBYAK_DEFINE_TYPE(72, tRAO_INT);
 
 #undef XBYAK_SPLIT_ID
 #undef XBYAK_DEFINE_TYPE
@@ -553,6 +554,7 @@ public:
 			if (EDX & (1U << 22)) type_ |= tAMX_BF16;
 			if (maxNumSubLeaves >= 1) {
 				getCpuidEx(7, 1, data);
+				if (EAX & (1U << 3)) type_ |= tRAO_INT;
 				if (EAX & (1U << 4)) type_ |= tAVX_VNNI;
 				if (type_ & tAVX512F) {
 					if (EAX & (1U << 5)) type_ |= tAVX512_BF16;